Machine Learning Models
Machine Learning Model Types in Seeq
Beginning with the release, Machine Learning tools can generate multiple supported outputs in a single execution. Previous releases were limited to a single output per tool execution.
Existing outputs remain associated with the tool instance that originally created them and can continue to be edited from that tool. New outputs created using the multi-output capability are associated with the new tool instance and can be edited from that tool.
Seeq supports two types of Machine Learning Models: Prediction models and Anomaly models.
Prediction models are used to predict values based on input signals and generate a new signal. They take one or more signals as input and produce a new signal as output. This model type is used to derive a signal based on the values of other signals.
Example: Predicting Temperature using Relative Humidity and Compressor Power signals.
Anomaly models are used to detect irregularities or abnormal patterns in signals. They take one or more signals as input, and their primary output is a condition based on a logical predicate (< 0 or > 0) applied to the predicted outlier value from the model. Optionally, they can also produce signals as output, representing anomaly scores or other measures.
Example: Detecting drift in the working condition of a system using multiple signals.